Repository navigation
Conversation
Co-authored-by: Lusoris <lusoris@pm.me>
- Annotate SCRIPT_PATH: Path and REPO_ROOT: Path in measure_quant_drop.py, ptq_dynamic.py, and ptq_static.py matching ADR-0681 conventions to reconcile silent-revert gate detection against d170ef8 without allowlisting. - Add focused regression coverage in test_ptq_cli_contracts.py for CLI help, import isolation, and argv handling contracts across PTQ and quantization scripts.
lusoris
disabled auto-merge
September 24, 2026 20:13
lusoris
changed the base branch from
master
to
train/pre-rc1-correctness-20260924
September 24, 2026 20:14
lusoris
added a commit
that referenced
this pull request
Sep 25, 2026
…ec A11) Restores the non-numeric column filter in ai/scripts/feature_correlation.py that was originally landed in commit 5fc7391 (#1552) and silently reverted in 384d97d. When processing parquets containing non-numeric columns (e.g. string metadata columns such as 'codec' or 'chug_orientation'), calling .to_numpy(dtype=np.float64) on candidate columns caused a runtime crash with ValueError: could not convert string to float. - ai/scripts/feature_correlation.py: filter candidate_cols using select_dtypes(include="number") before calling to_numpy(dtype=np.float64), and log any skipped non-numeric columns. - ai/tests/test_feature_correlation.py: add deterministic regression test test_corr_main_skips_non_numeric_columns verifying that string and metadata columns are skipped without crashing, omitted from the JSON report and Pearson keys, and that numeric features are retained. - docs/rebase-notes.md, docs/state.md, changelog.d/: record restoration of BUG-048 item A11 and update bug tracking state.
4 of 8 tasks
lusoris
pushed a commit
that referenced
this pull request
Oct 5, 2026
… defects on the known-upstream page The known-upstream page listed #1562 only. It now has a dated section for the GPU reports checked on 2026-10-05: - #1566 (fixed upstream by #1552): the CUDA motion kernel above 8 bits advanced a 16-bit pointer by the byte stride. Not affected: the fork's SAD kernel reads a row through a byte pointer (load_sample()), equal to the CPU in test_cuda_exact_twins, test_cuda_motion_tiny_frames and the depth and layout matrix; the planted upstream form fails above 8 bits and gives 26993 invalid reads under compute-sanitizer memcheck. - #1564 bugs 1 to 3 (CUDA masking border rows, per-warp rounding, x86 DWT last column) and its follow-ups: were affected, fixed, with the code and tests that hold each fix. docs/state.md records #1566 under "Confirmed not-affected".
lusoris
pushed a commit
to Tualua/vmafx
that referenced
this pull request
Oct 6, 2026
… defects on the known-upstream page (VMAFx#2178) * docs(upstream): record the CUDA 16-bit row-stride and integer ADM GPU defects on the known-upstream page The known-upstream page listed VMAFx#1562 only. It now has a dated section for the GPU reports checked on 2026-10-05: - VMAFx#1566 (fixed upstream by VMAFx#1552): the CUDA motion kernel above 8 bits advanced a 16-bit pointer by the byte stride. Not affected: the fork's SAD kernel reads a row through a byte pointer (load_sample()), equal to the CPU in test_cuda_exact_twins, test_cuda_motion_tiny_frames and the depth and layout matrix; the planted upstream form fails above 8 bits and gives 26993 invalid reads under compute-sanitizer memcheck. - VMAFx#1564 bugs 1 to 3 (CUDA masking border rows, per-warp rounding, x86 DWT last column) and its follow-ups: were affected, fixed, with the code and tests that hold each fix. docs/state.md records VMAFx#1566 under "Confirmed not-affected".
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Fixes the fail-closed pre-push mypy delta gate so branch and merge-base sources
are evaluated under the same checker configuration, with canonical module
identity and blocker-status handling. It also restores the accepted
ADR-0680/0681 shared bootstrap, parser, and replay-argv contract across twelve
current tiny-AI evaluation, quantization, and export scripts, guarded by an AST
red-cap and direct-file/module probes.
This closes
T-CI-MYPY-PREPUSH-BASELINE-USES-BASE-CONFIG-2026-09-21and the BUG-048AI-CLI helper cluster. BUG-048 itself remains open for its other independent
restoration sections.
Commit
d170ef86asilently replaced the already-accepted shared CLI/bootstraphelpers with hand-rolled path, parser, and argv handling. This PR selectively
restores ADR-0680/0681 behavior while preserving every later functional CLI
and provenance change; it does not roll back the training scaffold.
Type
fix— bug fixbuild/ci— tooling / infrafeat— new featureperf— performance improvementrefactor— no behavior changedocs— documentation onlytest— test-onlyport— cherry-pick from upstream Netflix/vmafsycl/cuda/simd— backend-specificChecklist
make verify-all, and thenormal pre-push gates are green.
ai/package passes: 1,327 passed, 1 skipped.Bug-status hygiene
docs/state.mdcloses both scoped rows with executable evidence andexplicitly keeps the wider BUG-048 restoration effort open.
Netflix golden-data gate
Local validation
Mypy gate repair
python3 scripts/git-hooks/test-pre-push-mypy.py: 32 passedAI CLI/bootstrap restoration
python -m--help: 24/24 passedai/package: 1,327 passed, 1 skippedRepository gates
make verify-all: passed, HISS 264 within the 276 baseline and all 29scanner-visible touched files clean
dedupe, formatting, mypy, and PR-body governance
findings before publication
Deep-dive deliverables
docs/research/2092-mypy-pre-push-config-baseline.mdanddocs/research/2088-bug048-ai-cli-helper-restoration.md.repairs and rejected suppression, whole-commit replay, and global
bootstrap-widening alternatives.
AGENTS.mdinvariant note —scripts/AGENTS.mdpins checker config,scope, canonical module identity, and fail-closed status handling;
ai/AGENTS.mdpins the twelve-script helper and Python 3.14 loader contracts.changelog.d/fixed/pre-push-mypy-config-baseline.mdandchangelog.d/fixed/bug048-ai-cli-helper-restoration.md.docs/rebase-notes.mdrecords both invariant groups.Reproducer
python3 scripts/git-hooks/test-pre-push-mypy.py PRE_COMMIT_TO_REF="$(git rev-parse HEAD)" python3 scripts/git-hooks/pre-push-mypy.py python3 -m pytest \ ai/tests/test_ai_cli_helper_restoration.py \ ai/tests/test_eval_report_run_provenance.py \ ai/tests/test_legacy_eval_report_run_provenance.py \ ai/tests/test_qat_smoke.py \ ai/tests/test_ptq_scripts.py \ ai/tests/test_ptq_cli_contracts.py \ ai/tests/test_measure_quant_drop_per_ep.py \ ai/tests/test_dnn_exporter_run_provenance.py -qExpected: 32 hook tests, no branch-introduced mypy findings, and 46 affected AI
tests pass.
Known follow-ups
Python Lintjob remains advisory and lacks a pinneddependency-rich type-check environment; that existing item remains tracked
as
T-CI-MYPY-JOB-CHECKS-NO-FILES-2026-09-21.scope until correctness and security cleanup completes.
branch as commit
862e1b08d; no admin bypass was used.